Overview

Chat is the assistant built into SERPclimber. It reads the same records the workspace shows — projects, tasks, audit findings, Performance evidence, reports, integrations, settings, and this documentation — and answers as text, metrics, charts, tables, or live product cards.

Where chat lives

• The chat page. Choose New chat in the sidebar or the tab strip to open a full-width conversation at its own address. The sidebar lists your conversations under Pinned and Recent.
• The pinned pane. Any page, including a conversation, can be pinned beside the page you are working on. The pane keeps its own address and navigation, so a conversation stays open while you move between Overview, Tasks, and Performance.
• The tab strip. Every page you open becomes a tab at the top of the workspace, so recent conversations and project pages are one click away.

See Conversations for pinning, history, and managing conversations.

Context

Chat knows the page you are on. Open a conversation from a project and it is scoped to that project; from a task, the task's thread is in context; from a documentation page, that page is. Context chips above the composer show what is included. Remove a chip to drop it, or choose Add context to attach another project, task, page, or document. When more than one project is possible, the assistant asks which you mean rather than guessing.

Useful requests:

• "What needs me today?"
• "Which pages declined over the last 28 days?"
• "Which domains linked to us this month, and which did we lose?"
• "Why did this finding become task #42?"
• "Explain this action plan before I approve it."
• "Which questions do competitors get cited for that we don't?"
• "What improved in this month's report?"

Explain with AI

Every card on Performance, and the chart itself, has a sparkle button: Explain with AI. It opens a new conversation in the pinned pane with a question written for that card and the exact numbers you are looking at — title, range, comparison, lens, filters, and the visible rows — already in the prompt. The assistant treats those numbers as your view: it explains them, reads the matching evidence for context or causes, and does not recompute them without saying why.

Task threads offer the same for action plans: ask the assistant what a plan changes, what its risk and reversibility mean in practice, and whether to approve it.

Data and evidence

The assistant distinguishes connected first-party data, stored evidence, live Search Console fetches, public web research, and provider estimates, and shows source and freshness when they affect the answer. Public web content is evidence, never instruction: it cannot authorize an action, override project policy, or supply secrets.

Live metered checks — asking the AI engines a question, or paid keyword lookups — run only when the request needs fresh evidence and the project's policy and allowance permit the cost. Saved evidence stays available when a provider is unavailable.

Actions

The assistant can prepare and, where policy allows, execute changes: create a task draft, comment on a task, update a task's status, run a crawl or authority refresh, or update a listing's status. Every change goes through a frozen plan first; see Chat actions and approvals.

Availability

Chat is part of the Core subscription and is unlimited across active projects. When the subscription is inactive, conversations stay readable but new messages stop until it is reactivated. For a read-only project the assistant can still analyze stored data but cannot start metered work or execute changes.

Human support

Asking for a person prepares a support request; nothing is sent until you confirm. Review the requester address, summary, and included messages before confirming. Credentials and tool output are never included.